The Winning Formula: Human Ingenuity Meets Software Development -Ari Kaplan

This talk explores the intersection of data, analytics, and artificial intelligence in sports and business. The speaker shares personal experiences, starting with his pioneering role in data analytics for professional baseball teams, drawing parallels to the cultural shifts highlighted in the movie Moneyball. He further discusses his work with Formula 1 teams and how real-time data impacts race strategies, including predicting weather changes around tracks. The speaker emphasizes the importance of feature engineering in maximizing insights from data, demonstrating its application in various sports and industries. He also touches on the potential of AI in detecting fraud, using the infamous Enron case as an example, and shares his vision of using data for humanitarian efforts in identifying wrongly imprisoned individuals.

So, if you were to think of movies, there would be two actual movies based on my life and five movies that described my life journey, all of which have to do with getting information, data, and AI. So, the first one is Moneyball. Quick quiz, how many have you seen or heard of the movie Moneyball? See? Not many others. Does anyone heard of Brad Pitt? Has anyone heard

of Taylor Swift? Okay, we're we're making progress. So, great movie, uh but it it's based partly on my experiences where I was one of the first five people to do uh data and analytics at a sports organization. Only five in the world at the time in the 1980s, probably before everyone, including you in this room, was born. So, how can you take information and better prepare your

players, better predict how much the players will perform in future years, but it was also a change in the culture, changing the way that teams operated. And there is some conflict in there. We're very much seeing the same thing with AI right now. Half of the conference here is educational content, like showing code, which I'll do on Friday, and half of it is like directional. We're losing

our jobs. We're not losing our jobs. So, same thing with Moneyball. Sports said you can't change the game. It's not numbers-driven. It's athletic, your heart-driven. Um then so so I I'm known in the industry for that movie, creating and leading analytics departments for many professional baseball teams, but I was also a scout evaluating the talent. Then I left after about 30 years in baseball and joined um

consulting, joined uh Formula 1. Has anyone heard of McLaren? Um so I worked traveling with a race strategy team with some of their drivers, Daniel Norris, Lando I'm sorry, Lando Norris, Daniel Ricardo. And then Brad Pitt makes a movie. I was on set during the film. So the joke is in 3 years Brad Pitt will be a movie about a Databricks evangelist. So we'll see. And then

um yeah, maybe just jump down to here and then I'll conclude there. BlackBerry. I am from Chicago. Um Now for the people joining, this is my second time in India. I was here 3 years Um I also went to Delhi, um Agra, Mumbai, and this trip I went to Jai Jaipur and Pushkar and then came here. but I'm from Chicago and the BlackBerry was one of the

biggest missed opportunities in business history. They owned the market of personal like mobile devices. And I started my own software company. I I worked at US Robotics. If anyone here heard of the Palm Pilot? They were the very first um before there was a mobile cell phone even, a digital assistant where you can stay connected to the internet from your phone. So I worked there and then

started my own software company on it. Tried to partner with BlackBerry and they wanted to close us out, keep it proprietary. But they could have should have been the next Apple. Um they Great movie based on that experience. So everyone's industry now is kind of at a tipping point. You could be the the BlackBerry where your company um it all depends goes down while others pop up.

Uh Anthropic, my favorite company of the moment and that had zero revenue 2 years ago and now they're one of the five most valuable private companies out there. If you want to learn how to fail, watch that movie. And then the other two uh I always encourage people to do beyond your work something positive for the world, whether it's doing good deeds or charity. And for me,

it's to use data and AI to find people around the world that are in prison for crimes they didn't commit or they're arrested by the government of another country. Try to get them out if they're innocent. Um so Bridge of Spies, Tom Hanks, was uh based on my partner. Um I'm one of uh four people who the Russians had given top security clearance to get into their

archives, their information of all their prisoners from 1947 to the present. Um millions of pages of documents. So, how do I use data and AI to find missing people? Raoul Wallenberg was a Swedish diplomat hero that rescued 80,000 Hungarian Jewish people during the Holocaust, but um he was kidnapped by the Soviets, disappeared in their gulag. So, in addition to Databricks, um I'm the president of the investigation

into what happened. So, I'm super excited. Anyone hear of Jamie Lee Curtis, the actress? Um she won the Oscar for Everything Everywhere All at Once movie. Um she, Brian Cox, Jake Gyllenhaal just filmed um a documentary based on on my analytic investigation. So, it should come out later this year. Pinch myself, seems like a dream, but um that that's part of my life. So, think I have

to be here for this button to work. Um and and like what some of the fun things in addition to doing analytics, I look at players from a human standpoint, and one of my biggest claim to fame was the star of the Chicago Cubs, probably no one here has heard of him, Anthony Rizzo, became a star with the Yankees, but I was the person to recommend the

Cubs acquire him based on analytics and him as a person, and the very second his hand caught the ball was the time the Cubs more than any sport, any team in the world, um had the longest drought of not winning a championship. 108 years. It's a lot. What was that the name? 108 names to Shiva's wife, right? That's the number of years of the Cubs not winning

the World Series, that and and they won finally. So, that was good. And then Taylor Swift, I I mentioned her name before, um people who don't like sports, I understand, um but I have worked not with her, with her agent, and with her boyfriend Travis Kelsey. If you go to LinkedIn, it's my pinned post on LinkedIn where I interview him saying, "Travis, you're won the Super Bowl.

Um how do you use data and AI?" And we talked for about 8 minutes on how he uses Well, it's really a team uses computer vision to find trends in his opponents. So, I do that to wake people up after lunch if you like Taylor. And then back to Formula 1, everyone here knows that many adventures worked with their race and worked with their drivers on either

making the car go faster or helping the perform even better. So, that that was And then, how do you get a pit stop? The car comes in, changes the tires, puts new gasoline in, the driver drinks, and they're out in under 2 seconds. And McLaren won the um the fastest ever. Car goes in 2 seconds, car goes out. And that can only be done with AI. Even

where the people changing the tires stand, how to get better equipment to push air into the tires. super data intensive environment. So, speaking of data intensive, I don't know what your company is. I've met people here. Uh Tesco, one of the sponsors, you know, retail, finance, healthcare. Um but for real-time, uh Formula 1, I can't think of anything faster. There are milliseconds that differ from winning the

race to not winning the race. And just the number of the algorithms for 80,000 car components is incredible. Um from a math standpoint, the number of combinations called 80,000 factorial. So, the number of combinations are greater than the number of atoms in the universe. So, it's impossible um maybe a quantum computer one day will do it. So, with AI, you can like eliminate obvious uh routes and

finally do these simulations. Um almost 400 million simulations run every single week. What compute power brings you that? So, that's a very very interesting AI challenge. And I did a fantasy Formula 1. I know betting's not allowed here, so it's all for fun here. But how do you pretend you're the race car driver, what tire Do you want a hard tire, soft tire? Uh one use case

was predicting the weather around the track to 5-minute increments. Where then the weather channel only changes the report once an hour, we needed every 5 minutes more accurate than the weather station would. We did a successful use case and we had an advantage over the other cars of our competition. And that was great. And I don't want to take any credit, but McLaren did very well last

year, actually winning the So, I talked baseball, I talked Formula 1, talked cricket, know my audience, more cricket fans than baseball fans. Um but AI is getting better and better. So, I've had the honor working with the athletes. I I don't view myself as athletic, but I would work with like super tall, the strongest people, smart people in the world. They liked me because the data that

I had would help them become a better player and help them understand how their competition as humans would do, what are their patterns. But now it's getting easier and easier. Um I heard Jason at this conference a lot. Take data, use AI to bring anybody down into 17 points. And then um it's just still math, but you need a human to take those points and make information

out of it. So, I was going to say this is my friend. I had another He's not my friend. The I I was over um the house of one of my partners in Jaipur, and his son um had a similar analyst just taking a video in high school, not a rich family or anything, and you could extract his arm is bent a little back. You want to

try to keep it straight. You start getting meaning out of these dots. What angle you want to have your angle parallel to the ground. This is kind of downward a little bit at the point of release. Like separation, how high? I like that form on on the tip. In baseball, that is huge. Your front foot plants forward or not helps you with your accuracy. And you could

see the differences in players. So now you have it's called feature engineering. A feature is a variable. You have a boring JSON file. If you look at it, it's just XYZ and time. And it it doesn't give you insights. This gives you more insights. Then you make a new feature on top of it, trunk angle. That that's the name that a human comes up with. Then you

have a feature on the feature on the feature, the change in the trunk angle or the trunk angle combined with forward movement velocity. That makes a new feature. With these new features of the data you already have, you can get much better insights and make the team winning. And this is applied to everything. So like in retail, you have what a customer is, but you could have

different types of customers. You could have velocity. Is the customer buying more and more, less and less, or straight and flat in their purchasing power. So you can get way way way more accurate insights on your data, whatever your industry is. Then this is basketball. Michael Jordan, anyone here is him? He's a I'm in Chicago. My favorite memory in sports was I had an office at the

Cubs stadium. It was from that wall to here and about from here to there and I had Michael Jordan and sky named Ernie Banks, just the three of us in my office talking about life and sports for for a while. That was my favorite moment. Um, but with basketball, you know, you have a dance of the data and the people using the data. So, you build more

features, like what is dribbling a ball, what is possession, um, in football or soccer as we call it, like Lionel Messi has different characteristics. How is his dribbling, his deception? And you make these terms and you can better align your your data with what the players are seeing. And a whole 'nother topic in this world of software development, you might call it metrics or you might call

it business definitions, or you might call it ontology, or you might call it metadata. That to me is one of the biggest things I hear around the world. Um, there's like 20 different mean standards, but if your business and you let your employees who are non-technical define these business terms, like a coach defining a shot, a heave, a jump, a layup, that gives them more actionable, practical

insights. And then here's a fun pretty thing. So, now that we have descriptions of how players move, you could define things like this person's laying down, leaning forward, following. You can make more features on features on features on features and do strategy. If this player is moving this way and that way, call it a forward rush. If, um, this player kicks it up in the air and

over, that's a passover. Um, and you can figure out like in in consumer product goods, it's called path to purchase. It's not that you made an advertisement and then someone buys it. It could be you made an advertisement, made a discount, gave out a coupon and then people buy. Three three path to And then getting external data, you have feature engineering. A lot of the roles of

people in this room, you have software development, you also have data engineers, data scientists, AI folks, collect external data, do an ETL process, merge data. For data science, the more types of data that you have, generally the better connected the insights are to the real world. So, this is for example, weather data. And cricket's the same thing with the wind blowing in versus out, the ball may

land an extra 20 m depending on the wind even though the batter's the same skill. All right. So, now and then and one more philosophy, so in data science, which to me is classical AI, which is making predictions, like you have a number, you're forecasting future sales on the past, that's traditional machine learning and you have classifications. Um that generally is a math formula. You have one

optimal price for your product, one answer. But with the world of AI, it's not as deterministic. You can have describe this uh PowerPoint, you can have 10 different answers, all of them correct. one thing I learned is there's not one math formula of perfect cricket swing. Same thing in baseball. You can have your hand up, hand down, and for the individual person, these are all correct. If

you you a scout and you saw someone holding a bat like that, you would fire them. But, that player was so successful, he's now the manager of the Chicago Cubs. So, my point is, there can be more than one correct answer. AI is a totally different um way. So, quick quiz. Now, I'm going to talk about career in about I saw the 10-minute sign about 4 minutes,

then I got to run through an example, and then if you kick me out, I'll be in the hallway to talk more. So, uh tongue-in-cheek, who's the agent? You have a banker, and you have a banker. And, of course, the answer is they're both agents. You, as, you know, the software world and the world of agentic systems, today you're probably working mostly with humans, but we're at

a tipping point where more people are more things are being generated through agents. I work at Databricks, and I'm not joking. I have screenshot later on. This past week was when more than we reached more than 50% of our software code is written by AI agents. Claude, I was speaking with to one of their lead software developers, already more than half of their code is written by

agents. So, you um you in whatever role you are, but especially software uh you know, you're going to learn to work with both of them. So, let me fast forward. Moneyball, this is uh I was not played by Brad Pitt, but this guy named Jonah Hill. I was the geek in the in that movie. Just want to point that out. Um and it changed the game. This

is the other This is one frustration is I came up with one use case that said, "Based on this batter and this pitcher, I stand here, and there is a um 65% chance of the ball coming to me. I stand here and there's a 10% chance. So, stand here, right? You have like six times as likely. It was the most obvious use case. We at the Cubs

implemented it. We saved 90 runs that year, which means we won about 10 more games, which is about 100 million US dollars. I would have thought today is Wednesday, April something, within 1 week, maybe 2 weeks, everyone in the game would be doing this, but no. It took 8 years for the teams to really adopt it in real um it's kind of interesting I hear around the

conference. Some things are changing super fast, but others you still need to have the same foundation, but it does change. And everything changes. This is a graph of the labor market. Um if you are into data entry, your job's probably more at risk. If you're looking to develop a quality application, you're probably going to be more valuable. And I just kind of looked at some things on

the flight over here, some cool things I could say. Uh there's always been trends where things get automated. One point, I was the president of the worldwide Oracle user group. 20 years ago, they said DBAs would be automated. Still hasn't happened fully. When you started in the 1960s, you would have assembly code. And then they would build compilers. That just shifted what people focused on and more

jobs opened up. Frameworks replacing boiler plates, cloud replacing server. So, we're just in one of those phases. Anthropic saying there won't be any more software engineer in 12 months, well, now they're hiring. I'll let you take a picture. And they're hiring 450 and and I don't know how much you make, but they're at least on the website 300, 400,000. And then here's the photograph of Genie Coode

over 50%. So let me do a quick story in the remaining time. Um has anyone heard of the company called Enron? It's an energy company. Awesome. There's should put that up there. Another movie. I think it was called The Biggest Men in the Room or something like that. The biggest financial failure um in in America. It was this uh company where they basically lied. It was like

a real company. They did stuff, but it was a scam. And so I had a team at Databricks say if they had AI back then, would it have been able to detect it? Would detect this fraud and not let the company go bankrupt. The short answer is yes. Um this went to court. So they the court made public. You can look at it now. All emails, all

financial data, everything like that. So we ingested it into Databricks and made our own chatbots and so on. This by the way goes to zero. People are still in prison after that. So I gave you the answer. Let's go through it. Oh yeah, there's the whistleblowers. So the first is that blue line is how much money and value they have in their portfolio. And this gray line

is the value of what they're investing in. So is going down, but the value that they reported to the public and shareholders was staying the same. That's and that's a magic trick. Um that alone shows something is very, very wrong with the finances. Um, so these are like screenshots in in our product. So then the next thing would be to ask, you know, looking at our portfolio

valuation, why is it flat, but the value of what we have dropped? And you can see um, you can see my cursor right here. That's where it flipped. So now we have the raw data. And you could see the um, this is our chatbot based on the data that we collected on our own proprietary data. It was on what's called a total return swap. These are financial

terms. It's a hedge. And then, so we said something is very wrong. So you could ask, what instrument like what financial property is used to hedge the portfolio, and what was the value in March when it went down? And it said, you know, here is the swap portfolio, something called Raptor. Hint, Raptor is a fraud. Um, it And how is it performing? Uh, what's the balance? What's

the trend? So you can start seeing that. Then you can see um, you know, there's an honest quantitative model and the rigged the fraudulent finance model. And you can see at the month that it switched, um, it went from market to hardcoded. Hardcoded means someone like fixed the number and it wasn't real. It was what the human did. I know people know what hardcoded here means. Um,

then you can just ask further questions. Who is responsible? Look at that, the office of the chief financial officer. That's awesome. It took years of court cases to figure that out. And then, since they released the emails, you what did the the operations people say. And it said, "I cannot support these valuations unless I look at the legal documents." So, we know who's innocent, and we know

the email right here, we know who's guilty. Yeah, so with that, I'm time I'll I'll I'll just say thank you. I'll leave this up here. These are my pillars of of wisdom um while the next speaker gets ready, but I'll be out in the hallway if you want to ask questions, despite what what might seem like 3 minutes, but now I'm done. I'll be back Friday. Thank

you all so much. >> [music]